Cleaned LightGBM documentation #2886

shmoradims · 2019-03-07T23:35:52Z

LightGBM API, trainers, boosters, and options documentation. Part of #2522.
Source of documentation are:

sfilipi · 2019-03-08T07:23:35Z

src/Microsoft.ML.LightGBM/LightGbmArguments.cs

+        /// Gradient boosting decision tree.
+        /// </summary>
+        /// <remarks>
+        /// For details, please see <a href="https://en.wikipedia.org/wiki/Gradient_boosting#Gradient_tree_boosting">here</a>.


replace with "Gradient_tree_boosting" #Resolved

done

In reply to: 263686400 [](ancestors = 263686400)

sfilipi · 2019-03-08T07:38:12Z

src/Microsoft.ML.LightGBM/LightGbmArguments.cs

+                /// </summary>
+                /// <value>
+                /// 0 means disable bagging. N means perform bagging at every N iterations.
+                /// To enable bagging, <see cref="SubsampleFraction"/> should also be set to a value less than 1.0.


should also be set to a value less than 1.0 [](start = 71, length = 43)

Just curious, where did this information come from? #Resolved

three sources:
*https://github.com/Microsoft/LightGBM/blob/master/docs/Parameters.rst
*https://lightgbm.readthedocs.io/en/latest/Parameters.html
*ML.NET source code

In reply to: 263689038 [](ancestors = 263689038)

sfilipi

sfilipi · 2019-03-08T07:39:18Z

src/Microsoft.ML.LightGBM/LightGbmCatalog.cs

        /// </summary>
        /// <param name="catalog">The <see cref="RegressionCatalog"/>.</param>
-        /// <param name="options">Advanced options to the algorithm.</param>
+        /// <param name="options">Trainer options.</param>


Trainer [](start = 34, length = 7)

either estimator or algorithm. I think the other catalogs use "Algorithm advanced options" #WontFix

I've been changing them to "Trainer Options" based on Rogan's comment on AveragedPercentron. I don't want to change back everything. Trainer sounds good enough, unless there's a strong objection.

In reply to: 263689283 [](ancestors = 263689283)

I think this is preferred to "advanced".

In reply to: 263905459 [](ancestors = 263905459,263689283)

revoking review

rogancarr · 2019-03-08T21:45:01Z

test/BaselineOutput/Common/EntryPoints/core_manifest.json

@@ -11360,7 +11360,7 @@
        {
          "Name": "CustomGains",
          "Type": "String",
-          "Desc": "Comma seperated list of gains associated to each relevance label.",
+          "Desc": "Comma separated list of gains associated to each relevance label.",


separated [](start = 25, length = 9)

Nice #Resolved

rogancarr · 2019-03-08T21:59:21Z

src/Microsoft.ML.LightGBM/doc.xml

+    -->
+    <member name="LightGBM_remarks">
+      <remarks>
+        LightGBM is an open source implementation of boosted trees. For implementation details, please see 


boosted trees [](start = 53, length = 13)

gradient boosted decision trees. (or Gradient Boosting Machine for GBM?) #Resolved

changed to "Gradient Boosting Decision Tree" based on Tom's paper title.

In reply to: 263944693 [](ancestors = 263944693)

rogancarr · 2019-03-08T22:00:54Z

src/Microsoft.ML.LightGBM/LightGbmArguments.cs

+    /// <remarks>
+    /// LightGBM is an external library that's integrated with ML.NET. For detailed information about the parameters
+    /// please see https://github.com/Microsoft/LightGBM/blob/master/docs/Parameters.rst.
+    /// </remarks>


Is there a version info anywhere that we could put here so that it's clear what version we have and it's auto-updated when we update LightGBM? #Pending

I don't think we have an auto-update mechanism for external repo version in docs. Right now it's at 2.2.3 which is the latest. I can't think of a way to force-update the docs here, if LightGBM version is included in text.

It's also easy to look at the LightGBM nuget version that comes with ML.NET nugets. That's the most reliable way to find the version I think.

In reply to: 263945106 [](ancestors = 263945106)

rogancarr · 2019-03-08T22:02:06Z

src/Microsoft.ML.LightGBM/doc.xml

+    <member name="LightGBM_remarks">
+      <remarks>
+        LightGBM is an open source implementation of boosted trees. For implementation details, please see 
+        <a href='https://lightgbm.readthedocs.io/en/latest/index.html'>LightGBM's official documentation</a>.


documentation [](start = 91, length = 13)

Also link to the paper with Guolin and Tom. #Resolved

good point.

In reply to: 263945392 [](ancestors = 263945392)

rogancarr · 2019-03-08T22:05:26Z

src/Microsoft.ML.LightGBM/LightGbmArguments.cs

        [Argument(ArgumentType.AtMostOnce, HelpText = "Rounds of early stopping, 0 will disable it.",
            ShortName = "es")]
        public int EarlyStoppingRound = 0;

-        [Argument(ArgumentType.AtMostOnce, HelpText = "Comma seperated list of gains associated to each relevance label.", ShortName = "gains")]
+        /// <summary>
+        /// Comma separated list of gains associated with each relevance label. Used only by <see cref="LightGbmRankingTrainer"/>.


a sep [](start = 16, length = 5)

hyphen #Resolved

done

In reply to: 263946259 [](ancestors = 263946259)

rogancarr · 2019-03-08T22:06:08Z

src/Microsoft.ML.LightGBM/LightGbmArguments.cs

        [Argument(ArgumentType.AtMostOnce, HelpText = "L2 Regularization for categorical split.")]
        [TlcModule.Range(Min = 0.0)]
        [TlcModule.SweepableDiscreteParam("CatL2", new object[] { 0.1, 0.5, 1, 5, 10 })]
        public double L2CategoricalRegularization = 10;

+        /// <summary>
+        /// The random seed for LightGBM to use.


The random seed for LightGBM to use. [](start = 12, length = 36)

Please specify what happens if the seed is not set by the user. #Resolved

done

In reply to: 263946446 [](ancestors = 263946446)

rogancarr · 2019-03-08T22:06:59Z

src/Microsoft.ML.LightGBM/LightGbmCatalog.cs

@@ -14,20 +14,20 @@ namespace Microsoft.ML
    public static class LightGbmExtensions
    {
        /// <summary>
-        /// Predict a target using a decision tree regression model trained with the <see cref="LightGbmRegressorTrainer"/>.
+        /// Predict a target using a boosted decision tree regression model trained with the <see cref="LightGbmRegressorTrainer"/>.


a boosted decision tree re [](start = 34, length = 27)

"gradient boosting decision tree" to be consistent with other docs. #Resolved

done

In reply to: 263946628 [](ancestors = 263946628)

rogancarr · 2019-03-08T22:07:38Z

src/Microsoft.ML.LightGBM/LightGbmCatalog.cs

@@ -46,14 +46,14 @@ public static class LightGbmExtensions
        }

        /// <summary>
-        /// Predict a target using a decision tree regression model trained with the <see cref="LightGbmRegressorTrainer"/>.
+        /// Predict a target using a boosted decision tree regression model trained with the <see cref="LightGbmRegressorTrainer"/> and advanced options.


oosted decision tree regression [](start = 38, length = 31)

Ditto #Resolved

done

In reply to: 263946785 [](ancestors = 263946785)

rogancarr

🕐

revoking review

rogancarr

🍇 📚 🎰

(closest emojis to GBM i could find)

codecov · 2019-03-08T22:30:43Z

Codecov Report

❗ No coverage uploaded for pull request base (master@f5b9e43). Click here to learn what that means.
The diff coverage is n/a.

@@            Coverage Diff            @@
##             master    #2886   +/-   ##
=========================================
  Coverage          ?   71.81%           
=========================================
  Files             ?      812           
  Lines             ?   142644           
  Branches          ?    16090           
=========================================
  Hits              ?   102433           
  Misses            ?    35827           
  Partials          ?     4384

Flag	Coverage Δ
#Debug	`71.81% <ø> (?)`
#production	`67.95% <ø> (?)`
#test	`86.24% <ø> (?)`

Impacted Files	Coverage Δ
src/Microsoft.ML.LightGBM/LightGbmBinaryTrainer.cs	`83.52% <ø> (ø)`
...Microsoft.ML.LightGBM/LightGbmMulticlassTrainer.cs	`87.94% <ø> (ø)`
...Microsoft.ML.LightGBM/LightGbmRegressionTrainer.cs	`74.35% <ø> (ø)`
src/Microsoft.ML.LightGBM/LightGbmArguments.cs	`89.63% <ø> (ø)`
src/Microsoft.ML.LightGBM/LightGbmCatalog.cs	`100% <ø> (ø)`
...rc/Microsoft.ML.LightGBM/LightGbmRankingTrainer.cs	`73.46% <ø> (ø)`

zeahmed · 2019-03-08T23:27:55Z

src/Microsoft.ML.LightGBM/LightGbmArguments.cs

                [Argument(ArgumentType.AtMostOnce,
                    HelpText = "Minimum sum of instance weight(hessian) needed in a child. If the tree partition step results in a leaf " +
                        "node with the sum of instance weight less than min_child_weight, then the building process will give up further partitioning. In linear regression mode, " +
                        "this simply corresponds to minimum number of instances needed to be in each node. The larger, the more conservative the algorithm will be.")]
                [TlcModule.Range(Min = 0.0)]
                public double MinimumChildWeight = 0.1;

+                /// <summary>
+                /// The frequency of performing subsamplig (bagging).


subsamplig [](start = 48, length = 10)

subsampling.
#Resolved

good catch.

In reply to: 263963937 [](ancestors = 263963937)

zeahmed

Overall LGTM. I expect the description/comments for each item was taken from LightGBM documentation. So, I am not putting comments around that. Let me know if you need that.

LightGBM API, trainers, boosters, and options documentation.

91a8d21

shmoradims requested review from singlis, sfilipi, zeahmed and rogancarr March 7, 2019 23:35

shmoradims mentioned this pull request Mar 7, 2019

Docs and samples for the API reference site (P0 & P1 Trainers) #2522

Closed

sfilipi reviewed Mar 8, 2019

View reviewed changes

sfilipi previously approved these changes Mar 8, 2019

View reviewed changes

sfilipi reviewed Mar 8, 2019

View reviewed changes

Addressed PR comments and fixed test.

44d7022

rogancarr reviewed Mar 8, 2019

View reviewed changes

rogancarr previously requested changes Mar 8, 2019

View reviewed changes

rogancarr approved these changes Mar 8, 2019

View reviewed changes

zeahmed reviewed Mar 8, 2019

View reviewed changes

zeahmed approved these changes Mar 8, 2019

View reviewed changes

Addressed PR comments.

6f6c255

shmoradims merged commit 97c067c into dotnet:master Mar 11, 2019

shmoradims deleted the lightgbm_docs2 branch March 11, 2019 20:51

shauheen added this to the 0319 milestone Mar 12, 2019

ghost locked as resolved and limited conversation to collaborators Mar 23, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Cleaned LightGBM documentation #2886

Cleaned LightGBM documentation #2886

shmoradims commented Mar 7, 2019 •

edited

Loading

sfilipi Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 8, 2019

sfilipi Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 8, 2019

sfilipi left a comment

sfilipi Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 8, 2019

rogancarr Mar 8, 2019

rogancarr Mar 8, 2019 •

edited

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

rogancarr left a comment

rogancarr left a comment •

edited

Loading

codecov bot commented Mar 8, 2019 •

edited

Loading

zeahmed Mar 8, 2019 •

edited by shmoradims

Loading

shmoradims Mar 11, 2019

zeahmed left a comment

Cleaned LightGBM documentation #2886

Cleaned LightGBM documentation #2886

Conversation

shmoradims commented Mar 7, 2019 • edited Loading

sfilipi Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sfilipi Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sfilipi left a comment

Choose a reason for hiding this comment

sfilipi Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited Loading

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rogancarr left a comment

Choose a reason for hiding this comment

rogancarr left a comment • edited Loading

Choose a reason for hiding this comment

codecov bot commented Mar 8, 2019 • edited Loading

Codecov Report

zeahmed Mar 8, 2019 • edited by shmoradims Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

zeahmed left a comment

Choose a reason for hiding this comment

shmoradims commented Mar 7, 2019 •

edited

Loading

sfilipi Mar 8, 2019 •

edited by shmoradims

Loading

sfilipi Mar 8, 2019 •

edited by shmoradims

Loading

sfilipi Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr Mar 8, 2019 •

edited by shmoradims

Loading

rogancarr left a comment •

edited

Loading

codecov bot commented Mar 8, 2019 •

edited

Loading

zeahmed Mar 8, 2019 •

edited by shmoradims

Loading